Helicase-like transcription factor (Hltf)-deletion activates Hmgb1-Rage axis and granzyme A-mediated killing of pancreatic beta cells resulting in neonatal lethality

Epigenetic silencing of HLTF in human beta cells alters insulin secretion and is associated with diabetes. Here we examine the effects of Hltf-deletion on beta cell function in genetically engineered mouse models. The developmental timeline for the exclusive expression of full-length Hltf mRNA and protein overlaps with the organization of murine pancreatic islets. Reduced insulin in the pancreatic beta cells of global Hltf-deleted near-term (E18.5) fetal mice predicts postpartum hypoinsulinemia. Seventy-five percent of newborn global Hltf-deleted postprandial mice die at a ratio of 3:2 males to females with negligible serum insulin levels, and their apoptotic beta cells are devoid of insulin. They are also hypoglycemic. Newborn beta cell-specific Hltf-deleted mice succumb to the same loss of glucose homeostasis indicating the phenotype is solely attributable to loss of beta cell function. Confirmation that an intact immune system is an absolute requirement for the phenotype resulted from breeding the Hltf-deletion into the Rag2-IL2-null background. Triple null (Hltf-/-Rag2-/-IL2-/-) newborn mice that lack functional receptors for IL-2,-4,-7,-9,-15, and -21, and have severe lymphocyte developmental impairment (deficient T and B cells, no NK cells) are euglycemic and normoinsulinemic with normal survival rates. Transcriptomic profiling (RNA-seq) eliminated beta cell-specific Hltf-deletion on gene programs essential for normal development, and beta cell secretion. Significantly, Hltf-deletion induced IL33/beta cell signaling that promoted islet infiltration of fetal CD8+T cells and beta cell apoptosis. Our data suggest, Hltf-deleted newborns that are not profoundly hypoglycemic survive in the absence of memory CD8+T cells. These Hltf-deletion studies provide mechanistic insights to how Hltf-deletion in beta cells promotes their immune destruction. We previously used RNA-seq in conjunction with 3SEQ/transcriptome to quantify expression levels in brain (14), heart (15), and placenta (7). RNase levels in pancreas are 181,000-fold higher than brain (https://www.thermofisher.com/us/en/home/references/ambion-tech-support/nuclease-enzymes/tech-notes/rnase-activity-in-mouse-tissue.html). As a result, isolation of total RNA was highly variable. Total RNA (n=47 samples) was isolated, its integrity and purity were assessed (Agilent Bioanalyzer). Ultimately pancreata (n=6 samples; 3 test (Hltf rIPCre fl/fl with low (<15 mg/dL) blood sugar, and 3 Hltf +/+ euglycemic controls) perfused in situ with RNAlater (21) using a 1 CC syringe and a 26g 1/2 needle, and snap frozen were suitable (Table 2) for rRNA-depletion. cDNA was generated from rRNA-depleted samples and subjected to Illumina library preparation. Libraries were sequenced utilizing Illumina sequencing technology. Paired-end 100 nucleotide reads were aligned to genomic assembly mm10 and analyzed using the platform provided by DNAnexus, Inc. (Mountain View, CA) to generate an unbiased gene expression analysis report of RNA-seq; alternative splicing analysis of Hltf; mutation/RNA-editing analysis and parallel comparison of expression profiles between beta cell-specific Hltf-deleted pancreata and control pancreata. The power in detecting alternative splicing was dramatically increased by paired-end sequencing relative to single-end sequencing. FPKM (fragments per kilobase of transcript per million mapped reads) were mapped against mm10 with Tophat (V1.3.3) to obtain .bam mapping files that were input into Cufflinks for transcript assembly. Cuffdiff (V 1.3.0), part of the Cufflinks package, used the alignment reads for rigorous statistical comparison of two conditions (beta cell-specific Hltf-deleted and wild type control pancreata) and 3 biological replicates for each condition. The depth of sequencing was a minimum of 20 million sequencing reads per sample [90% Power, 5% significance level: 91+/- 4% of all annotated genes are sequenced at a frequency of 0.1 times/103 bases X 3 x 109 bases/sequencing read x 3 samples = 9 x104 reads/gene]. Data were imported into iPathwayGuide (Advaita Corporation) a next-generation pathway analysis tool. Standard enrichment parameters (log fold change, logFC = 0.6, p<0.05) were used.人β细胞(beta cell)中HLTF(HLTF)的表观遗传沉默会改变胰岛素分泌，并与糖尿病相关。本研究借助基因工程小鼠模型，探究Hltf基因敲除(Hltf-deletion)对β细胞功能的影响。 全长Hltf mRNA与蛋白的特异性表达发育时序，与小鼠胰岛(murine pancreatic islets)的构建过程高度重合。 全身Hltf敲除的孕晚期（胚胎期18.5，E18.5）胎鼠的胰腺β细胞中胰岛素含量降低，这预示着产后会出现低胰岛素血症。 75%的新生全身Hltf敲除餐后小鼠死亡，雌雄死亡比例为3:2；这些小鼠血清胰岛素水平几乎检测不到，其凋亡的β细胞内完全缺乏胰岛素，且均处于低血糖状态。 特异性敲除β细胞中Hltf的新生小鼠同样会出现葡萄糖稳态失衡，表明该表型完全源于β细胞功能丧失。 通过将Hltf敲除等位基因导入Rag2-IL2基因敲除背景小鼠中，证实完整的免疫系统是该表型出现的绝对必要条件。 三重敲除小鼠（Hltf-/-Rag2-/-IL2-/-）缺乏IL-2、-4、-7、-9、-15及-21的功能性受体，且存在严重的淋巴细胞发育缺陷（T细胞、B细胞缺失，无自然杀伤细胞）；这类新生小鼠血糖正常、胰岛素水平正常，存活率也处于正常范围。 转录组测序（RNA-seq）分析显示，β细胞特异性Hltf敲除会改变正常发育及β细胞分泌功能所必需的基因表达程序。 尤为重要的是，Hltf敲除会激活IL33/β细胞信号通路，促进胎鼠胰岛被CD8+T细胞浸润，并诱导β细胞凋亡。 本研究数据表明，未出现严重低血糖的Hltf敲除新生小鼠，在缺乏记忆性CD8+T细胞的情况下可存活。 这些针对Hltf敲除的研究，阐明了β细胞中Hltf缺失如何促进其被免疫系统清除的分子机制。 本团队此前曾结合RNA-seq与3SEQ/转录组技术，对脑（14）、心脏（15）及胎盘（7）中的基因表达水平进行定量分析。 胰腺中的核糖核酸酶（RNase，Ribonuclease）活性是大脑的181000倍（来源：https://www.thermofisher.com/us/en/home/references/ambion-tech-support/nuclease-enzymes/tech-notes/rnase-activity-in-mouse-tissue.html）。 因此，总RNA的提取稳定性极差。 本研究共提取47份样本的总RNA，并通过安捷伦生物分析仪（Agilent Bioanalyzer）评估其完整性与纯度。 最终，共6份胰腺样本符合核糖体RNA（rRNA，ribosomal RNA）去除实验的要求：其中3份为实验组样本（Hltf rIPCre fl/fl小鼠，血糖<15mg/dL），另外3份为野生型对照样本（Hltf +/+，血糖正常）；这些样本均使用1mL注射器与26G 1/2针头，以RNAlater（21）进行原位灌注后快速冷冻（详见表2）。 从经rRNA去除处理的样本中合成互补DNA（cDNA，complementary DNA），并进行Illumina测序文库构建。 使用Illumina测序技术对文库进行测序。 将双端100碱基读长序列比对至小鼠基因组组装版本mm10，并借助DNAnexus公司（美国加利福尼亚州山景城）提供的分析平台，完成RNA-seq的无偏基因表达分析、Hltf基因可变剪接分析、突变/RNA编辑分析，以及β细胞特异性Hltf敲除胰腺与对照胰腺的表达谱平行对比分析。 相较于单端测序，双端测序可大幅提升可变剪接事件的检测效能。 使用Tophat软件（V1.3.3）将FPKM（每百万比对读长的每千碱基转录本片段数，fragments per kilobase of transcript per million mapped reads）比对至mm10基因组，得到.bam格式比对文件，再将其导入Cufflinks软件进行转录本组装。 作为Cufflinks软件包的组成部分，Cuffdiff（V1.3.0）利用比对得到的读长序列，对两组样本（β细胞特异性Hltf敲除胰腺与野生型对照胰腺）及每组3个生物学重复样本进行严格的统计学差异分析。 每个样本的测序深度至少为2000万条读长序列[置信度90%，显著性水平5%：所有注释基因中，91±4%的基因可按照0.1次/103碱基的测序频率被检测到，计算公式为0.1次/103碱基 × 3×109碱基/测序读长 ×3样本=9×104 reads/基因]。 将数据导入下一代通路分析工具iPathwayGuide（Advaita公司）中。 分析采用标准富集分析参数：对数倍数变化（logFC，log fold change）≥0.6，p值<0.05。